Comparison of fully and semi-automated area-based methods for measuring mammographic density and predicting breast cancer risk

NATURE PUBLISHING GROUP 2014-04-02

BACKGROUND: Mammographic density is a strong risk factor for breast cancer but the lack of valid fully automated methods for quantifying it has precluded its use in clinical and screening settings. We compared the performance of a recently developed automated approach, based on the public domain ImageJ programme, to the well-established semi-automated Cumulus method. METHODS: We undertook a case-control study within the intervention arm of the Age Trial, in which ∼54,000 British women were offered annual mammography at ages 40-49 years. A total of 299 breast cancer cases diagnosed during follow-up and 422 matched (on screening centre, date of birth and dates of screenings) controls were included. Medio-lateral oblique (MLO) images taken closest to age 41 and at least one year before the index case's diagnosis were digitised for each participant. Cumulus readings were performed in the left MLO and ImageJ-based readings in both left and right MLOs. Conditional logistic regression was used to examine density-breast cancer associations. RESULTS: The association between density readings taken from one single MLO and breast cancer risk was weaker for the ImageJ-based method than for Cumulus (age-body mass index-adjusted odds ratio (OR) per one s.d. increase in percent density (95% CI): 1.52 (1.24-1.86) and 1.61 (1.33-1.94), respectively). The ImageJ-based density-cancer association strengthened when the mean of left-right MLO readings was used: OR=1.61 (1.31-1.98). CONCLUSIONS: The mean of left-right MLO readings yielded by the ImageJ-based method was as strong a predictor of risk as Cumulus readings from a single MLO image. The ImageJ-based method, using the mean of two measurements, is a valid automated alternative to Cumulus for measuring density in analogue films.

Open access content. Open access content
https://creativecommons.org/licenses/by-nc-sa/4.0
CC BY-NC-SA

UMV oai:jupiter.its.unimelb.edu.au:11343/263277
Sovio, U., Li, J., Aitken, Z., Humphreys, K., Czene, K., Moss, S., Hall, P., McCormack, V. & dos-Santos-Silva, I. (2014). Comparison of fully and semi-automated area-based methods for measuring mammographic density and predicting breast cancer risk. BRITISH JOURNAL OF CANCER, 110 (7), pp.1908-1916. https://doi.org/10.1038/bjc.2014.82.
0007-0920
10.1038/bjc.2014.82
1532-1827
1315695761

UNIV OF MELBOURNE
From OAIster®, provided by the OCLC Cooperative.

edsoai.on1315695761